Preply, one of the largest online language learning marketplaces, has shared the results of Lesson Insights, a learning support feature built on the OpenAI API. The system analyzes each one-on-one lesson and automatically generates personalized feedback on grammar and pronunciation along with tailored homework. According to the company, 75 percent of English-language learners use the feature, which holds a 4.7 out of 5 satisfaction rating[1].

Lesson Insights Delivers an AI Report Minutes After Each Lesson

Preply is one of the world's largest online language learning marketplaces, connecting more than 100,000 tutors with learners in over 180 countries through one-on-one lessons in more than 90 languages[1]. While language learning is built on human elements such as conversation, motivation, and cultural understanding, tutors are often burdened with repetitive administrative work like writing study plans and lesson notes, and learners want a tangible sense of whether they are actually improving[1].

To address this, Preply built Lesson Insights on the OpenAI API. Lessons take place in the company's online classroom, Preply Classroom, and are recorded and transcribed with the learner's consent. Insight generation is scheduled a few minutes before the session ends, so feedback is ready as the lesson wraps up, and a structured report arrives in the learner's and tutor's chat thread within minutes[1].

Each report includes a summary of key lesson topics, grammar corrections with explanations, vocabulary highlights and translations, pronunciation feedback, and recommended next steps[1]. The insights also feed directly into Preply's self-learning exercise engine, which automatically generates personalized homework based on the lesson[1]. The design keeps everything connected, from the human lesson itself through review and the next round of practice.

75 Percent of Learners Use It, and Keep Using It a Year Later

The published numbers suggest Lesson Insights has staying power beyond initial novelty. Some 75 percent of English-language learners actively use it, and more than 70 percent of tutors do as well. Around 75 percent of active learners continue engaging with the feature more than a year after adoption, and it holds a 4.7 out of 5 satisfaction rating based on more than 300,000 ratings submitted on the platform. Its product market fit score of 70 percent is well above the threshold often associated with strong customer demand[1].

The benefits for tutors are concrete. Tutor Michelle Garcia Ramos says she used to spend hours preparing classes and creating homework, but that time has been cut by more than half since she started using Lesson Insights. Beyond efficiency, the AI suggests ideas and highlights what each student should focus on next, making it easier to improve the lessons themselves[1].

AI Inside the Company Too: 94 Percent of Engineers Use Codex

Preply has not limited AI to customer-facing features. The company rolled out ChatGPT Enterprise across the organization, running enablement sessions for more than 600 employees in New York, Kyiv, London, and Barcelona, which lifted weekly active usage from 60 percent to 95 percent[1].

On the engineering side, around 94 percent of Preply's engineers use Codex and other AI coding assistants for code generation, pull request reviews, and debugging. Co-founder and CTO Dmytro Voloshyn explains that Codex lets the team write code at a previously unseen speed, freeing engineers to focus more on overall architecture and solving customers' problems[1]. Custom GPTs such as a Brand Voice GPT, which helps teams create content consistent with the company's tone, are also embedded in daily workflows[1].

Summary

What stands out in Preply's case is the strict division of labor: AI is not there to replace human tutors but to absorb repetitive administrative work so humans can focus on motivation and conversation. Lesson Insights connects lesson transcripts, feedback, and homework generation automatically, and the fact that 75 percent of learners keep using it backs up its value[1]. Discussions of AI in language learning tend to fixate on replacing teachers, but this approach, where AI builds the review cycle around human lessons, looks like a natural fit for a discipline where so many learners give up midway.

Source: https://openai.com/index/preply